Public beta

Built to make your local knowledge extraordinarily useful. Alfred is the AI-first workspace for everything you keep.

Alfred connects files, notes, projects, and answers into a local graph you can inspect before AI takes action.

Download for macOS Install with Homebrew

Ask about this workspace

Summarize recent imports, inspect projects, search memory, or draft knowledge artifacts.

Summarize recent importsCompare two projectsSearch memoryCreate a note from latest imports

Ask Alfred about this workspace

Knowledge graph

Understands your entire workspace.

Alfred builds a knowledge graph across imported documents, project notes, concepts, retrieval traces, and Dream maintenance so every answer can point back to evidence.

Document ingestion

Ingest every document without losing the original.

Drop Markdown, PDFs, images, JSON, YAML, CSV, or HTML. Alfred preserves the raw source, extracts structure, chunks evidence, and writes the graph only when the pipeline is traceable.

Open the ingestion pipeline

Inbox

Personal

Vision available

Local importVision available

Import documents

Drop Markdown, text, PDF, image, JSON, YAML, CSV, or HTML files here to send them through extraction, triage, content persistence, and knowledge sync.

Drag and drop files here

The active provider can inspect images and rendered PDF pages during import.

Queue

Import triage workbench for incoming documents.

2 items

logo-flame.png

Rendered 2/4 pages, running vision extraction.

624 KBProcessing

raw-sources.md

Content stored, extracting retrieval context...

18 KBDone

Multisearch

Search with vector, FTS, lexical, and time signals.

Ask one question and Alfred blends semantic recall, full-text search, lexical matching, and temporal context so the answer can cite the right source instead of the nearest embedding.

Explore multisearch

MCP long horizon

Use Alfred as a real MCP server for long work.

Connect Codex, Claude Code, or any MCP client to Alfred's local bridge. A session can recall prior turns, search memory, inspect documents, and resume the same line of work with citations.

See MCP context

alfred-desktopMCP HTTP

mcpinitialize alfred-desktop @ http://127.0.0.1:51175/mcp

ok15 tools: recall, search, documents.list, documents.get_detail

youwhere did we leave the board update?

calltools/call recall({ query: "board update risks", projectName: "Memory" })

mem3 turns from mcp.save_interaction · repo alfred

calltools/call documents.list({ workspaceId: "personal", query: "board update" })

okBoard Update.md, Q2 Risks.pdf, Support Export.csv

calltools/call documents.get_detail({ documentId: "board-update" })

ansResume with churn risk, support signal, and two source citations.

Dream

Let the workspace improve while you are away.

Dream watches retrieval quality, concept drift, stale projects, and graph candidates, then keeps routine maintenance automatic while logging risky or ambiguous changes.

Open Dream autopilot

Works locally

Secure, local, and inspectable.

Alfred is public, but intentionally early. The first release focuses on macOS Apple Silicon, local workspace storage, and provider choices that keep inference under your control.

macOS Apple Silicon beta
Local workspace storage
OpenAI, Anthropic, Ollama, and Embedded providers
Automatic Dream maintenance with review for risky changes

macOS beta

Build your local graph.

Download the latest signed macOS DMG from GitHub Releases, or install through the Homebrew cask.

brew install --cask alfred-labs/alfred/alfred-app

Release notes